Send a risk and see every program that wants it, how likely each is to quote, and why. Carriers, MGAs and MGUs publish appetite once. Every underwriting decision makes the next answer sharper.
Early access. The matching engine is being built now. Join the waitlist to get in first.
What a match looks like
Residential roofer, Florida, $2.4M revenue, general liability
- Coastal Trades GLMGA lane86% · In appetite
- Contractor SelectCarrier lane74% · In appetite
- Southeast RoofersMGU lane52% · Likely referral
- Residential BuildersCarrier laneOut · rule §4.2
Illustrative example · quick match 10 gas ($0.10)
One network, three front doors.
Agents find markets. Program owners fill their queue with risks they actually want. Developers put both inside the tools people already use.
Program owners
MGAs, MGUs, program managers and carriers. Claim your program, confirm the rules we draft from your guidelines, and work a referral queue with reason codes.
For programs →Retail agents
Stop emailing five wholesalers to learn that three of them don't write roofing. Find markets for a risk in seconds and pay per submission.
For agents →Developers and partners
Headless first. A REST API, an MCP server and a one-tag embed widget, billed per call to your organization.
For developers →Four cuts.
Send the risk
Structured JSON or a form, or the ACORDs, loss runs and emails you already have, extracted into the same shape.
Rules, then the model
Each program's written appetite runs first. A rule decline is final and names the rule. Everything else gets a calibrated likelihood and reasons.
Work the deal
Questions, subjectivities, counter-offers and declines stay in one thread. Every decline carries a reason code.
The network learns
Decisions and reasons retrain each program's model and sharpen the questions agents are asked up front.
Pay for the answer, not the seat.
Two meters, both in gas, where 1 gas = $0.01. Agents and developers pay per call. Program owners license the console. Recording an underwriting decision is always free, because decisions are what the network learns from.
Agents and developers
Per call, from your organization’s gas wallet.
| Quick match · shop before you submit | 10 gas |
|---|---|
| Structured submission · no AI in the path | 25 gas |
| Documents · AI extraction, then submit | 40 + AI extraction |
| Thread, status and answers | free |
Program owners
Carriers, MGAs, MGUs and program managers.
- Tiered by program size
- Program console and referral queue
- Rules drafted from your guidelines, plus rule suggestions
- Discount when you opt in to pooled learning
- Recording decisions is always free
Why structured is cheapest
A structured submission goes straight to the deterministic engine. Documents are extracted into the same shape first, and the extraction is billed at the AI engine’s own gas.
A typical ACORD and loss-run packet is expected to land around $0.75–$2.00.
Facts confirmed by the people who touch them.
Decisions verified by every party
Agent, wholesaler, MGA and underwriter each confirm the step they own. The next step in the thread checks the last.
Facts checked against the record
Loss runs and production are verified against records the other parties hold, not just what one side reports.
Pooled learning is opt-in
By default a program learns only from its own decisions. Opt in to share anonymized features and get the stronger network model back.
Your decisions stay yours
Decision data belongs to the program owner. Every score is explainable and versioned, and a person owns every decline of a referred risk.
Open to every system. Owned by none of them.
An appetite network is only as useful as the number of places it reaches. MarketsNinja stays neutral: any agency management system, any portal, any rater, any carrier. No exclusivity, ever.
- Agency management systems
- Agency portals
- Comparative raters
- MarketsNinja
- Carriers
- MGAs
- MGUs
- Wholesalers
- AI agents via MCP
Plug in through the API, the MCP server or one script tag.
Most volume never opens our site.
MarketsNinja is built API-first. Put appetite search inside an agency management system, a portal, a rater or an AI agent.
What you get Coming
- REST API and an MCP server exposing the same tools
- One-tag embed widget for agency sites and portals
- Signed webhooks: matched, question, quoted, declined, bound
- Per-call gas billing to your organization
The shape of a match
Send a structured risk, get ranked programs with a likelihood and reasons.
POST /v1/match
{ "class": "roofing-residential",
"state": "FL", "revenue": 2400000,
"lines": ["GL"] }
→ [{ "program": "coastal-trades-gl",
"likelihood": 0.86,
"reasons": ["class in appetite", "revenue in band"] }]Illustrative. The API opens with early access.
Frequently asked questions
What is MarketsNinja?
MarketsNinja is an independent insurance appetite network. An agent sends a commercial risk and gets back the programs that want it, how likely each is to quote, and the reasons. Carriers, MGAs and MGUs publish and manage their program appetite in one place.
Is it live?
MarketsNinja is in early access. The matching engine is being built now. Join the waitlist and we will invite you as agents, programs and developers come on.
How much does it cost?
Pricing is per call in gas, where 1 gas is $0.01. A quick match is 10 gas ($0.10). A structured submission is 25 gas ($0.25). A document submission is 40 gas plus the AI extraction gas, which the extraction engine bills directly. Program owners license the console per program, with tiered pricing and a discount for opting in to pooled learning.
Why is structured data cheaper than documents?
A structured submission goes straight to the deterministic engine with no AI in the path. Documents such as ACORD forms, loss runs and emails are first extracted into the same structured shape, and the extraction is billed separately by the extraction engine at its own gas.
Who owns the underwriting decisions?
The program owner. Decision data belongs to the program. By default a program learns only from its own decisions; pooled learning across programs is opt-in. A person confirms every decline of a referred risk.
Can I use it from my own system?
Yes. MarketsNinja is built API-first: a REST API, an MCP server for AI agents, and a one-tag embed widget for agency sites and portals. Calls are billed per call to your organization’s gas wallet.
Is MarketsNinja tied to one distribution platform?
No. MarketsNinja is a neutral network open to any agency management system, portal, rater, carrier, MGA or wholesaler. No exclusivity.
Throw us your hardest risk.
Tell us who you are and we will bring you in as the engine opens. Early agents get a free gas balance to start.
We only use your email to invite you to early access.